Unit Conversion Pricing Errors: How I Fixed Silent Quote Bugs (Simply Explained)

A Quote That Looked Perfect and Cost $440 Anyway

I worked with a shade manufacturer that pulled pricing from six different supplier catalogs. About 9,500 rows of stuff. Tubes, fabric, hardware, brackets, motors, all of it.

Their system took those rows, calculated the costs, and printed quotes that looked clean enough to send a customer without a second glance.

That was the problem.

The quotes looked perfect. The math added up. The PDFs looked professional. But a lot of them were quietly wrong. Sometimes they overcharged the customer. Sometimes they lost money on their own jobs. Nobody knew which.

These are the most dangerous bugs you can have. A quote that throws an error gets caught. Someone sees the red warning and fixes it. But a quote that prints cleanly and is wrong by 11 percent goes straight to the customer, gets accepted, and becomes a signed contract. The mistake never raises its hand.

Why It Broke (and Why It Would Keep Breaking)

The system made one simple assumption. Fabric is always priced by the square foot. Tubes are always priced by the foot. Hardware is always priced by the unit.

Clean. Consistent. And wrong.

Suppliers don't care about your assumptions. They sell things the way their own warehouse and accounting work. So the six catalogs didn't agree with each other.

Some mixed metric and imperial measurements. Some sold tubes by the carton, others by the foot. One vendor sold hardware in pairs, another in singles. And one column, the most important one, was labeled "price per square foot" when the supplier actually meant "price per square yard."

A square yard is nine times bigger than a square foot. So the system read that price as nine times cheaper than it really was. After all the other math, that landed as an 11 percent error on those line items.

Eleven percent on a $4,000 quote is $440 gone. Multiply that across hundreds of quotes and you see why the books stop adding up at year end.

Here's the part people miss. This isn't a problem you fix once by cleaning the spreadsheet. New supplier prices arrive every month. Catalogs change. Vendors switch their packaging. You could hand-fix all 9,500 rows today and be wrong again in thirty days.

The fix has to live inside the import process itself. Automatic. Every time. With a paper trail you can actually trust.

How I Built a Pricing System You Can Prove Is Right

I rebuilt the way supplier data gets brought in, using three separate layers. Think of it like a kitchen with three stations.

Station one keeps the raw ingredients exactly as the supplier sent them. Every weird column, every odd value, untouched. You never edit it. This is your receipt. When a quote looks wrong six months later, this is where you go to see what the supplier actually charged.

Station two records the recipe. For every row, it writes down the original unit (per square yard), the original price, the conversion that got applied (multiply by nine), and a confidence score on whether it got the unit right. This is the layer almost nobody builds, and it's the one that matters most. It shows your work.

Station three serves the finished dish. These are the clean, ready-to-use prices the quoting engine actually reads. It never touches the raw data directly. It just gets prices in the units it expects and does its job.

The payoff is simple. You can trace any number on any quote back to the original supplier row and see exactly what conversion happened in between. If a conversion was wrong, you fix the recipe in one place and re-run it. You don't re-import anything. You don't call the supplier.

That's what "trustworthy" actually means. Not "we cleaned the data." It means "we can show our work on every single number."

Why I Use AI to Read the Units (Not the Labels)

The square-yard-labeled-as-square-foot disaster proves the whole point. The label lied. If your system trusts the column header, you've already lost.

So I use AI to figure out the real unit from context instead of the label. It compares the price against similar products from other suppliers. It looks at the patterns each vendor tends to follow. It reads the product description for clues.

When a fabric labeled "per square foot" is priced nine times higher than every other fabric in the data, the AI flags it. A human skimming a spreadsheet would never catch that. The AI catches it because it's comparing against everything else it has seen.

But here's the rule that keeps this safe. The AI never does the math. It just decides "this is probably per square yard, I'm 91 percent sure." Then plain, tested code does the multiplication. You never want AI doing arithmetic on your profit margins.

And when the AI isn't confident? That row doesn't get converted automatically. It goes to a human to confirm. High-confidence rows flow through. Uncertain ones wait for a person. That balance is the entire design.

I'll be honest. AI gets units wrong sometimes too. It isn't magic. That's exactly why the human review step exists.

The Trap Even the AI Couldn't See

There was a sneakier bug underneath the obvious one.

Fabric actually has two different prices hiding in one field. There's the area you sell to the customer (the square footage of the finished shade) and there's the way the supplier sells the roll (by length). These are not the same number. They're connected by how wide the roll is.

The system had quietly mashed both into one column, which leaked margin on every fabric line. Not random errors. A steady lean in one direction, which is the kind of slow drip that erodes profit without ever setting off an alarm.

The AI never would have caught this one. It can spot a price that's nine times too high. It doesn't know that a shade business sells in finished area but buys in roll length. That came from understanding the business, not the data.

That's the part people forget. The AI handles the volume and the pattern-spotting. The human knowledge sets the whole thing up in the first place.

How to Check Your Own Quotes This Week

Pull ten recent quotes. For each line, trace the cost back to the original supplier invoice. Not the cleaned version. The original.

If you can follow every number from the quote back to a source price and see exactly what conversion happened in between, you're in good shape. If you can't, you have no way to prove your quotes are right. Not "probably right." You literally cannot prove it.

I build these systems as the person running them, not as someone who hands you a slideshow and disappears. I live in this every day. My own DTC fashion brand in San Diego prices over 564 products automatically, and I've learned the hard way that the difference between a pricing system you trust and one you just hope is right comes down to one thing: being able to trace every number back to where it came from.

If you suspect your quotes are quietly off and you can't prove they aren't, that's the conversation worth having. Not a sales pitch. Just a look at a few of your own quotes to see what's really going on.

Ready to bring AI leadership into your company?

I work with a small number of companies at a time. If you're serious about AI, apply to work together and I'll review your application personally.

Apply to Work Together